Childhood and adulthood risk factors for socio-economic differentials in psychological distress: evidence from the 1958 British birth cohort

Social Science & Medicine 55 (2002) 1989–2004

Childhood and adulthood risk factors for socio-economic differentials in psychological distress: evidence from the 1958 British birth cohort C. Powera,*, S.A. Stansfeldb, S. Matthewsa, O. Manorc, S. Hopea a

Department of Paediatric Epidemiology and Biostatistics, Institute of Child Health, 30 Guilford Street, London, WC1N 1EH, UK b Department of Psychiatry, St Bartholomew’s and the Royal London School of Medicine and Dentistry, Queen Mary and Westfield College, London, UK c School of Public Health and Community Medicine, The Hebrew University and Hadassah, Jerusalem, Israel

Abstract Social inequalities in psychological status have been attributed to health selection and to social causation. We used data from the 1958 British birth cohort, followed over three decades, to identify causes of inequality in adulthood. Psychological status prior to labour market entry influenced inter-generational mobility, but selection effects were weaker for intra-generational mobility, between age 23 and 33. However, selection failed to account for social differences in risk of distress of approximately threefold in classes IV&V compared with I&II. Both childhood and adult life factors appeared to contribute to the development of inequalities. The principal childhood factors were ability at age 7 for both sexes and adverse environment (institutional care for men and low class for women). Adult life factors varied, with stronger effects for work factors (job strain and insecurity) for men and qualifications on leaving school, early child-bearing and financial hardship for women. Gradients in psychological distress reflect the cumulative effect of multiple adversities experienced from childhood. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Psychological distress; Social class; Social mobility; Health selection; Birth cohort; UK

Introduction Socio-economic differences have been observed across a range of mental disorders, including, schizophrenia, anti-social personality disorder and affective disorders, specifically depression and anxiety (Dohrenwend & Dohrenwend, 1974; Leaf, Weissman, Myers, Tischler, & Holzer, 1984; Dohrenwend, 1990; Bruce, Takeuchi, & Leaf, 1991; Dohrenwend et al., 1992; Kessler et al., 1994; Lewis et al., 1998), most studies showing higher rates of disorder among lower socio-economic groups. For more common mental disorder and distress the findings are less consistent (Dohrenwend & Dohrenwend, 1974; Weich & Lewis, 1998; Stansfeld, Head, & Marmot, *Corresponding author. Tel.: +44-20-7905-2106; fax: +4420-7242-2723. E-mail address: [email protected] (C. Power).

1998a), especially during early adulthood (Macintyre & West, 1991; Glendinning, Love, Hendry, & Shucksmith, 1992; Miech, Caspi, Moffitt, Wright, & Silva, 1999). In general, socio-economic status differences in health have been attributed to either health selection or to social causation. According to the selection hypothesis, men and women with pre-existing illness drift down the social scale and, conversely, those with better health tend to move up the scale. (Thus, health selection is also referred to as health-related social mobility.) Selection effects may be stronger for some conditions, such as schizophrenia and conduct disorder, than for other conditions, notably affective disorders (Dohrenwend et al., 1992; Miech et al., 1999). Likewise, the strength of selection effects may vary with life stage, with more pronounced effects at the time of labour market entry when the individual’s adult social position is established. It has been shown that the influence of psychological

0277-9536/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 9 5 3 6 ( 0 1 ) 0 0 3 2 5 - 2

1990

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

disorders on educational achievement commences from early in childhood (Offord et al., 1992) and continues in relation to adolescent conduct disorders (Jayakody, Danziger, & Kessler, 1998; Miech et al., 1999). In turn, education strongly influences adult social position (Caspi, Elder, & Herbener, 1990; Kuh, Head, Hardy, & Wadsworth, 1997). Psychological status may have an impact on adult social position not solely through education, but also through its effect on social functioning, skills and relationships. Thus, selection effects related to psychological status may be stronger at the transition from class of origin to own destination class (inter-generational mobility) than with any changes in social position during adult life (intra-generational mobility). This expectation of greater health selection during inter-generational mobility is not specific to psychological status, but is thought to apply to other health measures (West, 1991). Alternatively, social causation could explain health differences through the experience of adversity and stressors in low social status, and conversely, with more favourable conditions experienced in higher social groups. Social differences in factors relevant to mental health are evident at each life stage onwards from birth (Power & Matthews, 1997). First during early life, social causation may begin with conditions in the home environment that are regarded as central to the development of emotional well-being (Offord et al., 1992). A broad range of influences may be relevant, some of which are denoted by family structures, such as one-parent families, parental divorce and institutional care (Quinton & Rutter, 1988; Rutter, Quinton, & Hill, 1990); while other dimensions relate to the quality of family functioning, encompassing emotional support and stimulation, parental aspirations and involvement (Brooks-Gunn, Klebanov, Liaw, & Spiker, 1993). Early educational experiences may also contribute to the development of emotional well-being (Hertzman & Wiens, 1996), particularly through the development of self-esteem and mastery, and these will continue to play a role through to adolescence. Second, social causation may be linked to material disadvantage in childhood or in adult life (Lundberg, 1991; Offord et al., 1992; Weich & Lewis, 1998). Material circumstances could affect psychological status through social comparison, whereby deprived individuals are adversely affected because of perceived inequity (Wilkinson, 1997). Perceived inequity may be greater in societies with larger income inequalities, which in turn, may increase ill-health (Wilkinson, 1997), although this has been questioned (Muntaner & Lynch, 1999). Alternatively, material circumstances may affect other relevant factors, such as educational achievement, family structure and relationships, which in turn may increase the risk of poor mental health, although financial disadvantage may exacerbate or even

underlie these risks, for example, as experienced by lone mothers (Brown & Moran, 1997; Hope, Power, & Rodgers, 1999a). Thus, there may be either direct or indirect mechanisms linking material circumstances and psychological status. Third, social causation includes adult social relationships that impact on adult mental health and that vary by socio-economic status (Turner & Marino, 1994; Stansfeld et al., 1998a). These factors to some extent are a continuation of family functioning and structure (marital status, lone parenthood) in the early home environment, but in adulthood they also include social support and networks. It is well established that divorce and separation are associated with higher rates of psychiatric distress (Bloom, Asher, & White, 1978; Kitson & Morgan, 1990) as demonstrated in the life events literature (Paykel et al., 1969). For women, there may be an additional influence of the burden of childcare, as indexed in many studies by number of children, single motherhood and teenage pregnancy, all of which have been associated with adverse psychological distress (Brown & Harris, 1978; Weissman, Leaf, & Bruce, 1987; McLanahan & Adams, 1989; Maughan & Lindelow, 1997; Lipman, Offord, & Boyle, 1997; Weich, Slogett & Lewis, 1998). Beyond structural factors, quality of social relationships is predictive of mental health. In particular, lack of emotional support and negative aspects of close relationships are associated with higher rates of psychological distress (Oxman, Berkman, Kasl, Freeman, & Barrett, 1992; Stansfeld, Fuhrer, & Shipley, 1998b). Lack of emotional support may also be a vulnerability factor increasing the risk of depression in the face of life events (Brown & Harris, 1978). Fourth, factors related to labour force participation are part of social causation. There is evidence that unemployment and job insecurity adversely affect a range of mental health outcomes from minor distress to suicide (Banks & Jackson, 1982; Rodgers, 1991a; Burchell, 1994; Ferrie, Shipley, Marmot, Stansfeld, & Davey Smith, 1998; Gunnell et al., 1999). Other conditions at work, including high levels of job demands, low work social support and low decision latitude may influence psychological distress (LaRocco, House, & French, 1980; Wall et al., 1997; Stansfeld, Fuhrer, Shipley, & Marmot, 1999) and decline in psychological functioning (Martikainen, Stansfeld, Hemingway, & Marmot, 1999). Most aspects of labour force participation are strongly patterned by socioeconomic position, and in some instances appear to have a strong effect on the social gradient in depressive symptoms (Stansfeld et al., 1998a). Other factors may be involved in social causation, including, several health related behaviours, although the direction of association with adult mental health is not always well established. Co-morbidity has

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

been shown between alcohol problems and psychiatric disorder (Rodgers et al., 2000) and also, less consistently, between obesity and self esteem (French, Story, & Perry, 1995). Potentially, such factors might contribute to socio-economic differences in psychological status. Previous studies on socio-economic gradients in psychological status tend to focus on a limited range of explanations and hence, a comprehensive analysis of health selection and social causation is lacking. One exception is a study of psychological distress in the British 1958 birth cohort to age 23 years (Power, Manor, & Fox, 1991; Power & Manor, 1992). This study showed a threefold increase in psychological distress from the highest to lowest social classes in men and a five-fold difference for women. Health selection was evident, although inequalities in psychological distress were largely due to an accumulation of diverse factors from birth onwards rather than to health selection (Power et al., 1991; Power & Manor, 1992). Follow-up of the cohort to age 33 years has shown declining levels of distress among women (Rodgers, Pickles, Power, Collishaw, & Maughan, 1999), but persisting social gradients for both sexes. In the ten years since age 23 there has been a further cumulation of influences regarded as important for adult mental health. During this period, most individuals in Britain complete their education, gain employment, leave the family home and commence their own family. Some individuals had already divorced by age 33 and have experienced detrimental effects on their mental health, particularly for lone-parents (Hope, Rodgers, & Power, 1999b). Our broad purpose is to examine the role of health selection and social causation in the development of social inequalities in psychological distress in the 1958 birth cohort, followed to age 33. Specifically, we assess (i) whether psychological status influences the socio-economic destination that an individual achieves (a) at the time of the transition from social class of origin to own class and (b) within the period of early adulthood; (ii) the impact of health selection on social differences in psychological distress; and finally (iii) the role of social causation, as encompassing factors from early life through to recent years in early adulthood.

Methods Sample The 1958 birth cohort study comprises all children born in England, Scotland and Wales during one week in March 1958. Data were collected at birth, and at ages 7, 11, 16, 23 and 33. Immigrants to Britain born in the same week were added to the sample at 7, 11 and 16

1991

years. At age 23, 12,537 subjects were interviewed (76% of the target population), and at age 33, 11,405 subjects were interviewed (73% of the target population). As expected, sample attrition has been associated with under representation of those with the most disadvantaged backgrounds. Such biases tend to be small: 21.2% of men in the original sample had been born into classes IV&V compared with 20.5% who are included here in the analyses of health selection and 18.2% who had complete information for assessment of social causation (multivariate analysis); for women the figures are 21.4%, 20.7% and 19.0%, respectively. Thus, those remaining in the study are considered to be generally representative of the original birth cohort (Ferri, 1993). Data were obtained from parents, schools (teachers and doctors) and participants at ages 7, 11 and 16 years, and through personal interviews at ages 23 and 33 (Centre for Longitudinal Studies Institute of Education, 1994). Measures Psychological distress The Malaise Inventory was used to indicate psychological distress at ages 23 and 33. The Inventory is a selfcompletion 24 item checklist of symptoms of depression, anxiety and psychosomatic illness (Rutter, Tizard, & Whitmore, 1970) with acceptable internal consistency and validity in this study sample (Rodgers et al., 1999). It is scored by the summation of endorsed items, and totals were imputed for subjects with one or two items missing by determining their relevant percentile point on the population distribution of scores from the items which were responded to. As in previous work we used a cut-off score of 7 and above to identify those with a high level of psychological distress (Power & Matthews, 1997; Hope et al., 1999a). Social class Social class at age 33 years was based on current or most recent occupation and categorised using the British Registrar General’s classification. Four social groups were used: I&II (professional and managerial), IIInm (other non-manual), IIIm (skilled manual), and IV&V (unskilled manual). Both men and women were allocated to a social class on the basis of their own occupation. Most subjects (86% of men and 66% of women) had an occupation at age 33; an additional 11% of men and 28% of women were classified according to their most recent occupation, and only 3% of men and 6% of women could not be classified. Among women who were currently employed at 33 years, 32% worked part-time (o30 h/week). Similar social gradients in psychological distress exist for women in full-time and part-time jobs (Matthews, Power, & Stansfeld, 2001).

1992

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

Explanatory factors, from birth to age 16 years Inheritance and developmental factors Social class at birth was based on father’s occupation, and categorised as above for age 33, with the addition of a fifth category for ‘no male head of household’. In multivariate analysis class at birth was dichotomised into non-manual and manual groups with no male head of household included in the latter group. Walking unaided by 1.5 years and talking by age 2 were reported by mothers in 1964. Height at age 7 was measured by medical staff, to the nearest inch and converted into centimetres. Enuresis age 3–11 years was derived from the mother’s report of child’s bed wetting by day after 3 years and by night after 5 years. Educational factors Nursery attendance was based on mother’s report of pre-school experience (class, school or playgroup) or not. Frequency of parents’ reading to the child at age 7, was based on reported weekly reading habits (frequent, occasional or hardly ever). Ability: teacher’s assessment of oral, reading, creative ability, plus number work, and awareness of the world around. Categorised as average and above or below average and summed to create a continuous composite measure for age 7 ranging from 0 to 5. Receiving special educational treatment because of physical and mental disability at ages 7 and 11. Qualifications achieved by the end of school were categorised into four groups: none; up to 4 ‘O’ levels or equivalents; more than 4 ‘O’ levels or equivalents; and 2 or more ‘A’ levels. Family factors The number of family moves and the amount of time the cohort member had spent in care were recorded in 1964, both covering the period from birth to age 7. Alcoholism in the family when the cohort member was aged 7 was identified as a difficulty by the health visitor. Parental interest in the education of the cohort member was assessed by the teacher at ages 7, 11, 16. A score was derived (range 0–3) by adding the number of ages for which no parental interest (either parent) was identified. For multivariate analysis scores 2 and 3 were combined. Parental aspirations for the child’s education were ascertained at age 11 as the parent’s wishes that their child stay on at school and pursue further education or training. Parental divorce before age 16 was categorised separately from divorces after age 16 and parental deaths occurring before age 16. A third group comprised intact families. Whether the cohort member had ever been in care by age 16 (local authority or voluntary care) was reported at the 16-year survey.

Material circumstances Housing tenure age 11 was categorised into three groups: council tenants, owner-occupier and others (private rented, tied and other). In multivariate analysis two categories were used: council tenants and others. Household amenities at ages 7, 11, 16 are indicated by the sum of the three ages (range 0–3) at which the household shared basic amenities (use of bathroom, indoor lavatory, and hot water supply). Household crowding at ages 7, 11, 16 is indicated by the sum of the 3 years at which the household was overcrowded (i.e. more than 1.5 persons per room). Behavioural factors School absence, age 16 due to ill health in the year before interview as reported by parents and categorised as 1 month or more, and less than a month. Smoking, age 16 was obtained from self-reports of the number of cigarettes smoked in the past week, categorised as none (non-smokers) versus 1 or more. Behaviour at ages 7, 11, 16 was assessed with the Bristol Social Adjustment Guide (BSAG) at ages 7, 11 and Rutter school behaviour scale (Rutter, 1967) at age 16. The latter includes 26 items on behaviour and emotional status, six of which form an antisocial sub-scale (for example, ‘often destroys or damages own or others’ property’) and four items form an emotional sub-scale (for example, ‘often worries, worries about many things’); the 16 remaining items are not categorised. Total scores for the BSAG and Rutter scales were square-root transformed to improve the distribution of the scale and approximately the top 13% was categorised as poor socio-emotional adjustment or ‘‘problem’’ behaviour. The number of ages (7, 11 or 16 years) at which problem behaviour was identified were summed to derive a score ranging from 0 to 3. For multivariate analysis scores two and three were combined and the variable treated as continuous. Explanatory factors, from age 16 to 33 years Work characteristics Job insecurity at age 23 was identified from reported security in the subject’s current job (secure versus not secure). Unemployment by age 23 was the proportion of time (none, up to 7%, and more than 7%) that the subject was unemployed in the interval from when they were available for work to age 23. Redundancies between ages 23 and 33 were also reported and classified as none versus one or more redundancy. Unemployment between ages 23 and 33 is the proportion of time spent unemployed out of the total time available for work. Classification of unemployment among women was difficult because of the tendency for women to categorise themselves as housewives rather than unemployed. This factor was therefore included only for men. Psychosocial

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

(high strain) factors at age 33 are represented by four variables on work demands and level of control (learning opportunities, monotony, pace of work and flexibility of breaks) of the main paid or unpaid job, with a score of 2 or more negative responses indicating high job strain. Job insecurity at age 33 was identified from reports of the perceived future usefulness of work skills at that age. Health related behaviour Body mass index at ages 23 and 33 was calculated as weight in kilograms divided by height in meters squared and grouped as: o20.0, 20.0–24.9, 25.0–30.0 and >30.0. Alcohol consumption in the previous week at age 33 was grouped as (i) non/special occasion drinkers, (ii) light (0–10 units/week for men, 0–5 for women), (iii) moderate (11–35 units, 6–21 units), and (iv) heavy (>35 and >21 units) drinkers. Problem drinking was identified at age 33 from a score of 2 or more on the CAGE scale (Mayfield, McLeod, & Hall, 1974) in which respondents reported whether they had ever had one of four problems related to their drinking. Family structure and social support Marital status was reported at ages 23 and 33 as single, married, separated, divorced and widowed. Cohabitees were classified as single at age 23, but married at age 33. Social networks were assessed as the frequency of friends and family visits in the month before interview at age 23: 5 or more per week, 3–4 per week, 1–2 per week, 2–3 times in last month, once in last month, and none. Age at birth of the first child was categorised as by age 23 or later. Number of children by age 33 and lone parenthood status currently at age 33 were also reported. Social support at age 33 was based on emotional and practical support scales derived from the number of sources of support (Matthews, Stansfeld, & Power, 1999) and classified as having low emotional and practical support, high support on both these dimensions, with the remainder classified as intermediate. Material circumstances Financial hardship at age 23 is a derived score ranging from 0 to 4 indicating the number of current hardships, including: receipt of social security benefit, local authority housing tenure, no savings, and low (bottom 20%) net family income. Financial hardship at age 33 is a derived score ranging from 0 to 8, including: being in debt, no savings, local authority housing tenure, shared household facilities, mortgage/rent arrears, household dampness, access to a telephone, receiving low income social security benefits (Hope et al., 1999a). Subjects also reported whether they had ever been homeless by age 33.

1993

Data analysis Health selection Analyses were conducted to examine (i) whether psychological status influenced subsequent social mobility, and (ii) whether selection explains social differences in psychological distress. All analyses were performed separately for men and women. (1) To establish whether selection was operating intergenerationally, we assessed whether problem behaviour at age 16 was related to subsequent mobility. Thus, we used logistic regression models to calculate odds ratios of upward and downward mobility between ages 16 and 23 for those with problem behaviour at age 16. For health selection and intragenerational mobility we conducted parallel logistic regression analyses for mobility between ages 23 and 33, for those with psychological distress at age 23. (2) Logistic regression models were used to examine whether selection processes could account for social class differences in psychological distress at ages 23 and 33. Three logistic models were constructed to estimate the magnitude of class differentials in psychological distress. In the first, unadjusted model, the risk of psychological distress was estimated for each social class group, relative to social classes I&II. In the second model, the social gradient was adjusted for prior mental health (problem behaviour at age 16, or psychological distress at age 23) and social class of origin. In the third model, the social gradient was also adjusted for an interaction term representing health selection, i.e., prior health status by social destination, as recommended in previous modelling of selection effects (Lundberg, 1991; Rodgers & Mann, 1993). The 3-model sequence was conducted (i) for intergenerational selection, in which prior health was problem behaviour at age 16, social origins were represented by class at age 16, and social destination by class at age 23 (Fig. 2a); and (ii) for intragenerational selection, whereby prior health was psychological distress at age 23, origin class was at age 23, and destination class, at age 33 (Fig. 2b). Social causation In further analyses designed to assess the role of social causation, we conducted a series of logistic regression models to examine effects of factors from early life through to early adulthood. We identified potential explanatory factors from the literature and structured these into key conceptual areas, namely inheritance and developmental factors, educational and family factors, material circumstances and so on. We estimated odds ratios of psychological distress at age 33 for classes

1994

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

IV&V versus classes I&II, as an appropriate summary index for class differences in this population (Manor, Matthews, & Power, 1997). Initially, we fitted separate logistic models to estimate the effect of each explanatory variable on class differences in psychological distress, using the difference between the unadjusted and adjusted odds ratios to indicate the contribution of each variable. All models were for main effects only. Next for multivariate analysis, we selected variables that reduced the odds ratio for psychological distress by 10% or more. Variables were regrouped to overcome small sample biases and the additional problem of too few observations for multi-classification. Most variables were dichotomised to enhance the accuracy of the multivariate analysis (Harrell, Lee, Califf, Pryor, & Rosati, 1984) and to allow a greater number of variables to be investigated. The sample available for multivariate analysis was 2272 for men and 3068 for women. We repeated the univariate analyses for factors selected for the multivariate model, in order to identify whether any biases had been caused by the sample attrition. In general, the associations were similar, except for parental aspiration at age 11, which had a weaker effect in the multivariate sample. A potential problem in the multivariate analysis is collinearity arising from the correlation between the independent variables. For the final model we assessed the size of the problem and its impact on the estimated parameters and found that the results were not affected by collinearity (Wax, 1992). The multivariate analysis was constructed in stages following a temporal sequence: early life factors were analysed first, then childhood factors with those from early life, and so on. To some extent this approach allows for the timing of particular events and circumstances. However, to further clarify the temporal direction of effects we supplemented the final multivariate analysis with additional models designed to disentangle the direction of causation. These supplementary models were simple analyses as shown in Tables 2 and 3, but with an additional adjustment for prior or concurrent psychological status. If the effect of a specific factor remained after this adjustment, the factor was regarded as a predictor rather than an outcome of prior psychological status.

Results At age 33, 7% of men and 12% women had a high level of psychological distress, but as Fig. 1 shows, the prevalence of distress varied by social class. Among men, levels of distress ranged from 4% of those in professional and managerial classes (I&II) to 12% of those in unskilled manual classes (IV&V). Similarly for women, levels of distress ranged from 7% to 19%, respectively. This social class trend was significant for

20

Women Men

15

10

5

0 I &II

IIInm

IIIm

IV & V

Social class at age 33

Fig. 1. Social class gradient in psychological distress (%) at age 33.

both sexes (Mantel-Haenszel w2 49.5 for men, 101.2 for women). The social class gradient in psychological distress was examined for two groups, subjects with and without a current job at age 33. We found that the gradient was similar irrespective of whether social class was based on a current or most recent occupation. Health selection Table 1 shows the relationship between psychological status and subsequent inter-generational and intragenerational social mobility. Men and women with problem behaviour at age 16 were significantly more likely to be downwardly mobile by age 23 in comparison with others. Those with problem behaviour were also less likely to be upwardly mobile, with the exception of men in social class IIIN. In contrast, for intragenerational mobility there was less evidence of greater downward mobility among those with psychological distress at age 23; an elevated odds ratio for men originating in classes I&II was not statistically significant. Downward social drift of those with psychological distress was more apparent for women, particularly in those from social classes IIIN and IIIM. Men and women with elevated distress at age 23 were also less likely to be upwardly mobile, with the exception of women in social class IIIM. Because the social distribution differs for men and women, with more men in skilled manual occupations and proportionately more women in unskilled nonmanual occupations, the pattern of mobility varies by gender. We therefore repeated the health selection analyses combining classes IIIN and IIIM. The pattern of results was similar to that for the four social class groups. Then, in a further series of analyses we examined whether inter-generational social mobility was influenced by both conduct and emotional status, as indicated by the sub-scales on the behaviour scale. We

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

Men 5

1995

Women 5

Model 1 Model 2

4

Model 3

3

3

Odds ratio

Odds ratio

4

2

2 1

1

0

0 I&II

(a)

IIInm

IIIm

I&II

IV&V

IIInm

IIIm

IV&V

Social class age 23

Social class age 23

Model 1 = unadjusted odds ratio M odel 2 = adjusted for behaviour at age 16 and social class at age 16 M odel 3 = adjusted for behaviour at age 16, social class at age 16 and interaction between behaviour at age 16 & social class at age 23

Men

Women

5

4

3

Odds ratio

Odds ratio

4

(b)

5

Model 1 Model 2 Model 3

2

3

2

1

1

0

0

I&II

IIInm

IIIm

Social class age 23

IV&V

I&II

IIInm

IIIm

IV&V

Social class age 23

Model 1 = unadjusted odds ratio Model 2 = adjusted for psychological distress at age 23, social class age 23. Model 3 = adjusted for psychological distress age 23, social class age 23 and interaction between psychological distress & social class at age 33. Fig. 2. Health selection and the social gradient in psychological distress. (a) Effect of inter-generational mobility on the gradient at age 23. (b) Effect of intra-generational mobility on the gradient at age 33.

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

1996

Table 1 Psychological status and subsequent social mobility: odds ratios of upward and downward social mobility in those with poor emotional status relative to others (a) Intergenerational mobility and problem behaviour at age 16 Social class at age 16 Social mobility, ages 16–23 Downward Stable Upward n/N

Men (n ¼ 4131)

Women (n ¼ 4220)

I&II

IIIN

IIIM

IV&V

I&II

IIIN

IIIM

IV&V

5.63* 1.00 F 63/1063

14.48* 1.00 1.58 33/462

1.71* 1.00 0.43* 229/1808

F 1.00 0.38* 145/798

9.61* 1.00 F 37/1102

10.79* 1.00 0.87 24/447

1.77* 1.00 0.41* 155/1885

F 1.00 0.39* 103/786

(b) Intra-generational mobility and psychological distress at age 23 Social class at age 23 Social mobility, ages 23–33 Downward Stable Upward n/N

Men (n ¼ 4249)

Women (n ¼ 4482)

I&II

IIIN

IIIM

IV&V

I&II

IIIN

IIIM

IV&V

1.61 1.00 F 31/975

1.06 1.00 0.64 28/774

1.16 1.00 0.69 109/1740

F 1.00 0.74 70/760

1.23 1.00 F 77/1015

1.60* 1.00 0.73 290/2298

2.21* 1.00 1.84 78/390

F 1.00 0.87 178/779

n/N numbers with (a) problem behaviour at age 16/full sample and (b) psychological distress/full sample. *Odds ratio is significantly different from unity at 5% level.

found the same trend as the overall socio-emotional behaviour score at age 16: for both men and women, there was evidence of selection processes for both subscales, with greater downward and less upward mobility among those with high scores, although this pattern was more pronounced for conduct disorder. We next examined whether health selection could account for social class differences in psychological distress in early adulthood. Fig. 2 shows the social gradient in psychological distress at age 23. First, the unadjusted odds ratios of 3.64 and 4.24 (classes IV&V versus I&II) was significant for men and women, respectively (model 1 in Fig. 2a). Second, after adjustment for problem behaviour at age 16 and social class of origin (age 16), the gradient attenuated slightly, such that the odds ratio for classes IV&V reduced to 2.47 for men and 3.20 for women (model 2 in Fig. 2a). Third, the social gradient was adjusted also for an interaction between problem behaviour at age 16 and class of destination at age 23, representing health selection, with the result that the social gradient increased slightly, especially among men (model 3 in Fig. 2a). If health selection had generated inequalities in distress, then adjustment for the interaction representing selection would be expected to substantially reduce the odds of distress in classes IV&V versus classes I&II, but this effect is not observed. Similarly for intra-generational

mobility (Fig. 2b) there was no additional reduction in the social gradient in psychological distress at age 33 after adjustment for the interaction between psychological distress at age 23 and social class at age 33, representing health selection (model 3). For men, the odds ratio for classes IV&V was 2.84 without adjustment, reducing to 2.05 with adjustment for psychological distress and social class at age 23, but thereafter no further reduction was achieved with adjustment for the interaction representing health selection (odds ratio=2.07). For women the respective odds ratios are 2.88, 1.67 and 2.01. Hence, adjustment for intragenerational health selection widened rather than attenuated the social gradient for women (Fig. 2b). The odds ratio for classes IV&V versus I&II remained significant throughout these inter- and intra-generational analyses. To summarise, psychological status was related to subsequent social mobility, both inter- and intragenerationally, but we found no evidence that health selection contributed to the development of social gradients in psychological distress. The direction of the effect of adjustment for health selection suggested that social gradients in psychological distress might if anything be steeper in the absence of health selection. Given that health selection did not provide an explanation for the gradient in distress in this population, we considered

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

1997

Table 2 Odds ratios of psychological distress at age 33 (classes IV&V versus I&II) adjusted for childhood and adolescent factors Men

Women Adjusted

Reduction in OR-%

n

Unadjusteda

Adjusted

Reduction in OR-%

3.53* 2.99* 3.00* 3.02* 3.09*

3.24* 2.99* 2.99* 2.98* 2.98*

8 0 o1 1 4

5256 4967 4970 4646 4375

3.49* 3.76* 3.76* 3.85* 3.59*

3.12* 3.76* 3.74* 3.67* 3.48*

11 0 o1 5 3

4746 4766 4835 4143 4698

2.92* 2.97* 3.00* 3.22* 3.40*

2.95* 2.91* 2.21* 3.01* 2.48*

0 2 26 7 27

4945 4961 5058 4281 4906

3.79* 3.75* 3.62* 3.75* 3.70*

3.78* 3.74* 2.80* 3.62* 2.26*

o1 o1 23 4 39

Family factors Family movesFbirth to age 7 In care to age 7 Alcoholism in family at age 7 Parental aspirations, age 11 Parental interest, ages 7–16 Parental divorce by age 16 In care (ever) by age 16

4707 4778 4759 4649 5284 5392 3918

2.94* 3.04* 2.99* 3.21* 3.11* 2.98* 3.30*

2.95* 2.89* 2.99* 2.89* 2.72* 2.92* 2.90*

0 5 0 10 13 2 12

4917 4980 4954 4807 5480 5559 4108

3.81* 3.75* 3.81* 3.42* 3.68* 3.53* 3.41*

3.80* 3.71* 3.77* 3.04* 3.08* 3.43* 3.31*

o1 1 1 11 16 3 3

Material circumstances Housing tenure, age 11 Household amenities ages 7, 11, 16 Household crowding ages 7, 11, 16

4644 4811 4761

3.22* 3.10* 3.37*

3.08* 3.09* 3.37*

4 o1 0

4807 4986 4967

3.42* 3.58* 3.65*

3.03* 3.49* 3.61*

11 3 1

Behavioural factors School absence, age 16 Smoking, age 16 Behaviour score, ages 7–16

4008 4109 5390

3.03* 3.21* 3.16*

2.79* 2.95* 2.55*

8 8 19

4202 4318 5582

3.53* 3.33* 3.58*

3.25* 3.06* 3.11*

8 8 13

n

Unadjusted

Inheritance and developmental factors Social class at birth Walking unaided by 1.5 years Talking by 2 years Height, age 7 Enuresis, age 3–11

5059 4776 4772 4509 4187

Educational factors Nursery attendance Parent reads to child, age 7 AbilityF5 skills, age 7 Special education, ages 7–11 School qualifications, by 16y

a

a

Unadjusted odds ratios vary because of differences in sample size (n) for each variable. *95% CI excludes 1.

a range of other explanatory factors that might be involved in social causation. Social causation Potential childhood and adolescent explanatory factors for class differences in psychological distress are presented in Table 2. Odds ratios are given without adjustment, and with adjustments separately for each factor, in order to identify factors with the greatest influence on class differences in distress. Within the first group of inheritance and early developmental factors, class at birth had the greatest impact, with a reduction of 8% in the odds ratios for men (from 3.53 to 3.24) and of 11% for women (from 3.49 to 3.12), whilst other factors had a negligible effect. Among educational factors, the

strongest effects were seen for ability at age 7 and for qualifications obtained by the end of school odds ratios reduced by 26% for men and by 23% for women after adjusting for 7-year ability; and by 27% and 39%, respectively for end of school qualifications. Nursery attendance and parental reading had a negligible impact. Two family factors, namely parental interest and educational aspirations, also appeared to have an impact on class differences in distress for both sexes, with odds ratios reducing by 13% and 16%, respectively for men and women after adjustment for parental interest and by 10% and 11% for parental aspirations. A third family factor, being in care by age 16, also had an impact but only among men, reducing the odds ratio by 12%. In contrast, effects of material circumstances were generally weak, except that adjustment for housing

1998

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

Table 3 Odds ratios of psychological distress at age 33 (classes IV&V versus I&II) adjusted for adulthood factors Men

Women a

n

Unadjusteda

Adjusted

3 12 2 10 18 26

3314 4906 5398 N/a 5338 5322

2.51* 3.19* 3.23* N/a 3.51* 3.52*

2.48* 3.07* 3.24 N/a 3.18* 3.04*

1 3 0 N/a 9 14

2.78* 3.07* 2.83* 3.03*

o1 3 8 +o1

4890 5539 5646 5617

3.31* 3.61* 3.53* 3.55*

3.23* 3.57* 3.42* 3.80*

2 1 3 +7

2.99* 2.99* 3.28* 3.21* 3.32* 3.28* 2.86*

2.94* 2.93* 3.07* 3.09* 3.31* 3.27* 2.78*

2 2 6 4 o1 o1 3

4976 4975 5503 5412 5294 5480 5293

3.23* 3.23* 3.52* 3.37* 3.48* 3.48* 3.43*

3.01* 3.15* 3.44* 2.48* 3.28* 3.24* 3.30*

7 3 2 26 6 7 4

2.95* 3.04* 3.06*

2.55* 3.14* 2.02*

14 3 34

4975 5011 5646

3.23* 3.21* 3.52*

2.55* 3.22* 2.30*

21 +o1 35

n

Unadjusted

Adjusted

Working characteristics Job insecurity age 23 Unemployment age 23 Redundancies 23–33 Unemployment ages 23–33b Psychosocial strain, age 33 Job insecurity age 33

3978 4583 5409 5329 5036 5031

2.63* 2.94* 2.87* 3.26* 2.62* 2.62*

2.56* 2.60* 2.80* 2.92* 2.16* 1.94*

Health related behaviour Body Mass Index age 23 Body Mass Index age 33 Alcohol consumption, age 33 Problem drinking

4554 5372 5441 5441

2.79* 3.17* 3.08* 3.04*

Family structure/social support Marital status age 23 Social networks, age 23 Marital status age 33 Age at first child, by age 33 Number of children, age 33 Lone parenthood, age 33 Social support, age 33

4643 4642 5246 5154 4931 5243 4912

Material circumstances Financial hardship, age 23 Ever homeless to age 33 Financial hardship, age 33

4640 4691 5452

Reduction in OR-%

Reduction in OR-%

a

Unadjusted odds ratios vary because of differences in sample size (n) for each variable. Unemployment was not considered to be reliably recorded for women, due to their tendency to categorise themselves as housewives rather than unemployed. *95% CI excludes 1. b

tenure at age 11 reduced the odds ratio for women by 11%. Lastly among child-adolescent factors, there was a 19% reduction in the odds ratio for men after adjusting for socio-emotional behaviour and a 13% reduction for women, whilst other behaviours were associated with lesser effects. Table 3 presents the unadjusted and adjusted odds ratios for early adulthood factors. Several work characteristics had an impact on the odds ratio for men, in particular job insecurity at age 33 which was associated with a 26% reduction, but with psychosocial job strain and unemployment also reducing the odds ratio by 10% or more. For women, adjustment for job insecurity at age 33 reduced the odds ratio by 14%, but other factors had weaker effects. Neither of the health related behaviours examined, alcohol problems or BMI, substantially affected the odds ratio in psychological distress, and family structure and social support factors also showed modest effects. An exception was the age

that women had their first child, which was associated with a 26% reduction in odds ratio from 3.37 to 2.48. Finally, there was a substantial impact of material circumstances in early adulthood. Financial hardship at age 33 had the largest effect, similarly for both men and women in that the odds ratios reduced by more than a third, and hardship at age 23 also had a marked effect, especially for women, reducing the odds ratio by 21% (14% for men). Factors having the greatest effect on the odds ratio (i.e. associated with a reduction of 10% or more) were selected for multivariate analysis: 12 factors for men and 11 for women, as shown in Table 4. However, it was suspected that some factors could be outcomes of prior psychological status and so we attempted to disentangle the direction of causation, i.e. between predictors and outcomes, by repeating the analyses in Tables 2 and 3 with an additional adjustment for prior or concurrent psychological status. The impact of most factors was

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

1999

Table 4 Psychological distress at age 33 (classes IV&V versus I&II), adjusted for child, adolescent and adulthood factors Men ðn ¼ 2272Þ Unadjusted Adjusted for Social class at birth and Ability, age 7 and Housing tenure, age 11 and Parental aspirations, age 11 and Behaviour score, ages 7 to16 and Parental interest in education, ages 7–16 and In care by age 16 and School qualifications and Age at first child and Unemployment to age 23 and Financial hardship, age 23 and Unemployment, ages 23–33 and Psychosocial strain, age 33 and Job insecurity, age 33 and Financial hardship, age 33

3.12 F 2.32 F 2.36 2.11 2.14 2.08 2.10 F 1.89 1.84 1.80 1.58 1.41 1.35

unaffected, but modest reductions in the impact of some factors was observed suggesting that they were in part a marker of previous psychological status. Factors for men were: unemployment to age 23, psychosocial job strain at age 33, and financial hardship both at 23 and 33 years; and for women parental interest in education, hardship at age 23 and job insecurity. However, even for these factors, there was a substantial impact on the class differential above that associated with previous psychological status. We therefore included all selected factors in a multivariate analysis. When all factors were modelled simultaneously, it is notable that odds ratios for psychological distress decreased from 3.12 to 1.35 in men and from 3.67 to 1.43 in women. Thus, the overall effect of these factors was substantial, resulting in a non-significant odds ratio for both sexes. Analysed in the temporal sequence indicated in Table 4, it is evident that most factors contributed to the decrease in odds ratios, although some factors more so than others. In particular, the odds ratio reduced when ability at age 7 was included in the model, even for women with simultaneous adjustment for social class at birth. Thereafter the impact of specific factors appeared to vary for men and women. For men, there was an additional reduction in the odds ratio after adjustment for the child/adolescent behaviour score and further notable reductions were associated with 16–23 year unemployment, psychosocial job strain and job insecurity at age 33. For women, the strongest effects were seen for school qualifications and age at first child, and more modest effects for housing tenure at age 11, financial hardship at ages 23 and 33, and also for job insecurity. In comparison with univariate analyses

95% CI 1.81, 5.39 F 1.30, F 1.32, 1.17, 1.18, 1.14, 1.12, F 1.00, 0.97, 0.94, 0.82, 0.72, 0.69,

4.13 4.23 3.81 3.87 3.79 3.94 3.58 3.51 3.43 3.04 2.73 2.62

Women ðn ¼ 3068Þ 3.67 3.38 2.77 2.63 2.55 2.53 2.42 F 1.95 1.72 F 1.63 F F 1.54 1.43

95% CI 2.64, 5.11 2.42, 1.95, 1.85, 1.79, 1.77, 1.69, F 1.34, 1.18, F 1.12, F F 1.04, 0.97,

4.74 3.93 3.74 3.64 3.61 3.46 2.82 2.51 2.39

2.27 2.12

shown in Tables 2 and 3, effects were greatly diminished for adult factors, such as financial hardship for women. Nonetheless, adjustment for such factors still contributed to the reduction in odds ratio in psychological distress even after allowance for several earlier life factors. The odds ratio for men was no longer statistically significant after the inclusion of 23-year hardship level, but for women the odds ratio remained significant until adjustment for the final factor, hardship at age 33. Inevitably, the timing of some factors is uncertain and we therefore conducted supplementary analyses to assess whether the ordering of factors in the analysis affected the results. The impact of parental interest in, and aspirations for, their offspring’s education was affected by their inclusion in the analysis after housing tenure, but this was only evident for women because the impact of these factors among men had been affected by sample attrition. Hence, we were unable to disentangle the distinct effects of these factors. All 33-year factors contributed to the reduction of the odds ratio irrespective of the order in which they appeared in the analysis, although for men, the impact of job insecurity and psychosocial job strain was affected by order in which they appeared in the analysis.

Discussion In this large nationally representative population we found a social gradient in psychological distress in early adulthood, with greater levels of distress among men and women from lower social classes. Following the

2000

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

sequence of events, it emerges that psychological status influences an individual’s social destination, but for this cohort at least, this selection process is not a major factor in the development of the social gradient in distress. Other factors accumulating over the lifecourse, such as early life ability and adult work-related factors (for men) and age of first child-bearing and material circumstances (for women) suggest a predominant role for social causation. Clearly, these main findings are specific to a general measure of psychological distress (Malaise Inventory), which incorporates symptoms of anxiety and depression and somatic symptoms of emotional distress. Although the Inventory has been found to be a robust measure of distress with construct validity in early adulthood (Rodgers et al., 1999), nevertheless, it does not distinguish either between different types of affective disorder or between affective and personality disorders. Thus, our findings may not apply to all specific psychological disorders. A second consideration in the interpretation of our study is that when population samples are followed over decades sample attrition may cause bias in the results. We found sample bias for some potential explanatory factors (e.g. for parental interest in education), but relationships for most factors appeared to be unaffected by sample attrition. Set against these limitations our study has several strengths, primarily due to the prospectively recorded information. These data permit a detailed investigation of the relationship between factors at each life stage, and thereby provide an understanding of the temporal ordering involved. By testing explicitly the effect of explanatory factors on the socio-economic gradient we extend previous work on health selection, social causation and psychological distress in the 1958 cohort (Power et al., 1991). The role of health selection Within this cohort, as elsewhere (Rodgers & Mann, 1993; Timms, 1996), selection was seen to operate according to psychological status. Unlike most other studies (Power et al., 1991; Glendinning et al., 1992; Timms, 1996; Miech et al., 1999), we were able to differentiate between inter-generational and intra-generational selection and this revealed a stronger effect in association with inter-generational mobility. Two main explanations could account for this finding. First, the difference in effects may be largely artefactual, arising from the different measures of psychological status available in the study. Adolescent psychological status, prior to the transition into the labour market, was based on the Rutter school behaviour scale at age 16, whilst in early adulthood we used the Malaise Inventory at age 23. These measures have been shown to be related (Power et al., 1991), but nonetheless tap different aspects of psychological status. In particular, the 16-year

measure includes both antisocial behaviour and emotional state, while the 23-year measure focuses solely on emotional distress. Selection effects might be expected to be stronger for antisocial behaviour than for some other psychological states and we found some support for this argument within the 1958 cohort, as well as in previous studies (Miech et al., 1999). An alternative explanation for stronger inter-generational health selection is that the transition from class of origin to own class is the major life stage at which social sorting occurs, during which psychological and general health status can influence an individual’s career trajectory (Macintyre & West, 1991; Miech et al., 1999; Manor et al., 2002). Our findings were not, however, consistent with health selection playing a major role in the development of social class differences in psychological distress at age 23 or 33. If health selection were an important factor we would expect social differences in psychological distress to reduce after statistical adjustment for health-related social mobility, but there was no consistent evidence of this. On the contrary, the social differential increased rather than reduced, although the effects were inconsistent across sex and age. This pattern would appear to illustrate the proposition that selection can have variable effects on the relationship between socio-economic and psychiatric disorder (Goldman, 1994), and, in particular, that selection may act to limit rather than create social differences in health (Bartley & Plewis, 1997). It has been argued that given initial differences in levels of ill health between social groups, the upwardly mobile are likely to be healthier than stable members of the class of origin, but less healthy than members of the class of destination. In addition, the downwardly mobile are likely to be less healthy than stable members of the class of origin, but more healthy than members of the class of destination. Health selection could constrain socioeconomic gradients because the effect of these mobility trends is to increase the level of poor health in the highest class and reduce the level in the lowest class (Bartley & Plewis, 1997). It is interesting to note, therefore, that there is some suggestion of a constraining effect of selection here in the 1958 cohort. A further possibility is that mobility out of work reduces the effect of health selection on the gradient. When those with the poorest psychological health move out of employment rather than down the social scale, this may weaken the social gradient, and in addition, weaken the apparent effect of selection on the gradient (Bartley & Plewis, 1997; van de Mheen, Stronks, Schrijvers, & Mackenbach, 1999). In the present study we cannot discount the possibility that the effect of health selection is under-estimated, due to failure to consider movements in and out of work. However, it is likely that any under-estimation is minor because our social classification is based on current or most recent occupation. Thus most subjects (97% of men and 94%

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

of women) were classified to an occupational class and were therefore included in our analysis. Whilst acknowledging that inter-generational selection effects appear to be operating, we conclude nevertheless that the gradient in psychological distress is not due to selection effects. Clearly this applies specifically to the measure of distress used in the 1958 cohort and selection may emerge as more important for other psychological disorders (Dohrenwend et al., 1992; Miech et al., 1999). But for our investigation of psychological distress it was fortunate that we could also assess the role of social causation. Social causation As anticipated, both childhood and adulthood factors appeared to be involved in the development of the socioeconomic gradient in psychological distress. An effect of childhood factors was expected because, in this population, there are both continuities in psychological status (Power et al., 1991; Rodgers et al., 1999) and associations with socio-economic status (Power & Matthews, 1997). Continuities in psychological status within childhood (Verhulst & van der Ende, 1992) and from childhood to adulthood are widely recognised, with associations for antisocial behaviour (Robins, 1978; Tremblay et al., 1992), from childhood behaviour and personality to adult affective disorder, depression and other psychiatric disorders (Harrington, Fudge, Rutter, Pickles, & Hill, 1990; Robins & Price, 1991; Rodgers, 1991b). But, in addition to these child–adult continuities there appear to be other pathways through which childhood risks impact on socio-economic differences in adult psychological distress. One pre-eminent childhood factor emerging from our study is the individual’s ability at age 7. Ability is strongly related to both psychological distress in early adulthood and social class (data not presented) and thence makes an important contribution to social class differences in psychological distress for both sexes. We are not able to identify precisely what ability represents in this context, nor how it affects psychological distress in adult life. Nonetheless, it seems plausible that ability affects an individual’s sense of mastery and possibly their self-esteem, and in turn, influences the development of emotional wellbeing. The factors that affect ability at age 7, are better understood, at least in the 1958 cohort, as represented by stimulating home environment, parent involvement and aspirations for their child. Such factors appear to act through childhood ability in their effect on adult psychological distress, since they no longer show an effect after adjustment for ability at age 7. It is interesting that for men there was no additional impact of school qualifications on the socio-economic gradient, while for women, a strong effect of qualifications was evident and was only partially reduced by adjustment

2001

for childhood ability and other early life factors. It is not entirely clear why school qualifications is an important additional factor only among women, although it may be that qualification level is a stronger determinant of social position among women (Power & Manor, 1992). Other childhood factors that had an impact on the social class gradient in psychological distress in early adulthood were early socio-economic circumstances for women, and being in local authority care for men. Thus women may be particularly vulnerable to socio-economic adversity experienced throughout childhood, as suggested by previous work on this cohort, and moreover, the association between social origins and destinations appears to be stronger among women (Power & Manor, 1992). The impact of childhood adversity for men is evident from our finding that being in care contributes to social differences in psychological distress. This is consistent with other studies showing adverse adult psychological outcomes for institutionalreared men and women (Quinton & Rutter, 1988; Rutter et al., 1990). Social class gradients in psychological distress therefore appear in part to originate in childhood, but adult life factors are also relevant. Elsewhere we have shown that psychosocial job strain and job insecurity increases the risk of psychological distress among both men and women in this cohort (Matthews et al., 2001). Simultaneously, those work factors are strongly associated with social class and, consequently, as we show in the present analysis, job insecurity for both sexes and psychosocial job strain for men, appear to be important explanations in adult life for the social gradient in psychological distress. The influence of job insecurity and psychosocial job strain are less impressive after allowing for earlier life factors, but even so, notable effects were demonstrated in this cohort suggesting a unique contribution of factors in adult life. We cannot discount the possibility that the self-perceived basis of ‘high strain’ work and psychological distress may underlie relationships between the two factors and that the distinction between the measures may therefore be blurred. One argument against this is that, for women, job strain does not emerge as an important factor and, furthermore, it shows social patterns that are inconsistent with those for distress (Matthews, Hertzman, Ostry, & Power, 1998). Although there are few studies with which to make comparisons, our findings concur with one previous study showing an impact of decision authority and skill discretion on the social gradient in depressive symptoms (Stansfeld et al., 1998a). Surprisingly, we found here and elsewhere that adult family structure and burden of domestic roles had a minimal influence on the socioeconomic gradient in distress (Matthews et al., 2001). Two factors did, however appear to be, in general, important at least for women, notably age at first child and financial hardship in adulthood. Women who have

2002

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

a child at a young age may be at risk of a range of adverse social outcomes in early adulthood (Maughan & Lindelow, 1997). In turn these social destinations have implications for the living conditions and resources available to individuals. Lack of material resources is known to be associated with psychological distress (Pearlin & Johnson, 1977; Bruce et al., 1991; Brown & Moran, 1997; Stansfeld et al., 1998b) as shown in other work on this cohort (Hope et al., 1999a) and is clearly associated with social class. For women, at least the social gradient in distress appears to be influenced by the earlier child bearing and poorer material resources of those in lower classes. In conclusion, we have found that social gradients in psychological distress are primarily due to social causation. This finding is consistent with a previous claim that social causation is important for depression (Dohrenwend et al., 1992). Furthermore, we show that there is a cumulative deleterious effect of unfavourable childhood environment (being in institutional care for men and low social class for women) and poor ability at age 7, together with unfavourable adult circumstances, which appear to vary for men and women. This provides further evidence on the importance of multiple adversities in the aetiology of adult psychological disorders (Quinton & Rutter, 1988; Rutter et al., 1990; Rodgers, 1990). Our results suggest that interventions to reduce inequalities in adult psychological distress need to be targeted at both childhood and adulthood risk factors.

Acknowledgements This research was supported by a grant from the (UK) Economic and Social Research Council under the Health Variations Programme (L128251021) to Chris Power, Sharon Matthews, Stephen Stansfeld and Orly Manor. CP is a fellow with the Canadian Institute for Advanced Research.

References Banks, M. H., & Jackson, P. R. (1982). Unemployment and risk of minor psychiatric disorder in young people: Crosssectional and longitudinal evidence. Psychological Medicine, 12, 789–798. Bartley, M., & Plewis, I. (1997). Does health-selective mobility account for socioeconomic differences in health? Evidence from England and Wales, 1971 to 1991. Journal of Health and Social Behavior, 38, 376–386. Bloom, B. L., Asher, S. J., & White, S. W. (1978). Marital disruption as a stressor: A review and analysis. Psychological Bulletin, 85(4), 867–894. Brooks-Gunn, J., Klebanov, P. K., Liaw, F., & Spiker, D. (1993). Enhancing the development of low-birthweight,

premature infants: Changes in cognition and behavior over the first three years. Child Development, 64, 736–753. Brown, G. W., & Harris, T. O. (1978). Social origins of depressionA study of psychiatric disorder in women.. London: Tavistock. Brown, G. W., & Moran, P. M. (1997). Single mothers, poverty and depression. Psychological Medicine, 27, 21–33. Bruce, M. L., Takeuchi, D. T., & Leaf, P. J. (1991). Poverty and psychiatric status. Longitudinal evidence from the New Haven Epidemiologic Catchment Area study. Archives of General Psychiatry, 48(5), 470–474. Burchell, B. (1994). The effects of labour market position, job insecurity, and unemployment on psychological health. In D. Gallie, C. Marsh, & C. Vogler (Eds.), Social change and the experience of unemployment (pp. 188–212). Oxford: University Press Oxford. Caspi, A., Elder, G. H., & Herbener, E. S. (1990). Childhood personality and the prediction of life-course patterns. In L. Robins, & M. Rutter (Eds.), Straight and devious pathways from childhood to adulthood (pp. 13–35). Cambridge: Cambridge University Press. Centre for Longitudinal Studies Institute of Education, The Data Archive Distributor, Colchester, Essex. (1994). National child development study, sweeps one to five. National Birthday Trust Fund, National Children’s Bureau, City University SSRU [original data]. SN: 3148. Dohrenwend, B. P. (1990). Socioeconomic status (SES) and psychiatric disorders: are the issues still compelling? Social Psychiatry and Psychiatric Epidemiology, 25, 41–47. Dohrenwend, B. P., & Dohrenwend, B. S. (1974). Social and cultural influences on psychopathology. Annual Review of Psychology, 25, 417–452. Dohrenwend, B. P., Levav, I., Shrout, P. E., Schwartz, S., Naveh, G., Link, B. G., Skodol, A. E., & Stueve, A. (1992). Socioeconomic status and psychiatric disorders: the causation-selection issue. Science, 255, 946–952. Ferri, E. (Ed.) (1993) Life at 33: the fifth follow-up of the National Child Development Study. London: National Children’s Bureau. Ferrie, J. E., Shipley, M. J., Marmot, M. G., Stansfeld, S., & Davey Smith, G. (1998). The health effects of major organisational change and job insecurity. Social Science and Medicine, 46, 243–254. French, S. A., Story, M., & Perry, C. L. (1995). Self-esteem and obesity in children and adolescents: a literature review. Obesity Research, 3(5), 479–490. Glendinning, A., Love, J. G., Hendry, L. B., & Shucksmith, J. (1992). Adolescence and health inequalities: extensions to Macintrye and West. Social Science and Medicine, 35(5), 679–687. Goldman, N. (1994). Social factors and health: the causationselection issue revisited. Proceedings of the National Academy of Science, 91, 1251–1255. Gunnell, D., Lopatatzidis, A., Dorling, D., Wehner, H., Southall, H., & Frankel, S. (1999). Suicide and unemployment in young people. Analysis of trends in England and Wales, 1921–1995. British Journal of Psychiatry, 175, 263–270. Harrell, F. E., Lee, K. L., Califf, R. M., Pryor, D. B., & Rosati, R. A. (1984). Regression modelling strategies for improved prognostic prediction. Statistics in Medicine, 3, 143–152.

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004 Harrington, R., Fudge, H., Rutter, M., Pickles, A., & Hill, J. (1990). Adult outcomes of childhood and adolescent depressionFI. Psychiatric status. Archives of General Psychiatry, 47(5), 465–473. Hertzman, C., & Wiens, M. (1996). Child development and long-term outcomes: a population health perspective and summary of successful interventions. Social Science and Medicine, 43(7), 1083–1095. Hope, S., Power, C., & Rodgers, B. (1999a). Does financial hardship account for elevated psychological distress in lone mothers? Social Science and Medicine, 49, 1637–1649. Hope, S., Rodgers, B., & Power, C. (1999b). Marital status transitions and psychological distress: longitudinal evidence from a national population sample. Psychological Medicine, 29(2), 381–389. Jayakody, R., Danziger, S., & Kessler, R. C. (1998). Early onset psychiatric disorders and male socioeconomic status. Social Science Research, 27, 371–387. Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes, M., Eshleman, S., Wittchen, H. U., & Kendler, K. S. (1994). Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States. Results from the National Comorbidity Survey. Archives of General Psychiatry, 51(1), 8–19. Kitson, G. C., & Morgan, L. A. (1990). The multiple consequences of divorce: a decade review. Journal of Marriage and the Family, 52, 913–924. Kuh, D., Head, J., Hardy, R., & Wadsworth, M. (1997). The influence of education and family background on women’s earnings in midlife: evidence from a British national birth cohort study. British Journal of Sociology of Education, 18, 385–405. LaRocco, J. M., House, J. S., & French, J. R. (1980). Social support, occupational stress, and health. Journal of Health and Social Behavior, 21(3), 202–218. Leaf, P. J., Weissman, M. M., Myers, J. K., Tischler, G. L., & Holzer, C. E. (1984). Social factors related to psychiatric disorder: the Yale Epidemiologic Catchment Area study. Social Psychiatry, 19(2), 53–61. Lewis, G., Bebbington, P., Brugha, T., Farrell, M., Gill, B., Jenkins, R., & Meltzer, H. (1998). Socioeconomic status, standard of living, and neurotic disorder. Lancet, 352(9128), 605–609. Lipman, E. L., Offord, D. R., & Boyle, M. H. (1997). Single mothers in Ontario: sociodemographic, physical and mental health characteristics. Canadian Medical Association Journal, 156(5), 639–645. Lundberg, O. (1991). Childhood living conditions, health status, and social mobility: a contribution to the health selection debate. European Sociological Review, 7(2), 149–162. Macintyre, S., & West, P. (1991). Lack of class variation in health in adolescence: an artefact of an occupational measure of social class? Social Science and Medicine, 32, 395–402. Manor, O., Matthews, S., & Power, C. (1997). Comparing measures of health inequality. Social Science and Medicine, 45(5), 761–771. Manor, O., Matthews, S., Power, C. (2002). Health selection: the role of inter- and intra-generational mobility on social inequalities in health. (forthcoming).

2003

Martikainen, P., Stansfeld, S., Hemingway, H., & Marmot, M. (1999). Determinants of socioeconomic differences in change in physical and mental functioning. Social Science and Medicine, 49, 499–507. Matthews, S., Hertzman, C., Ostry, A., & Power, C. (1998). Gender, work roles and psychosocial work characteristics as determinants of health. Social Science and Medicine, 46(11), 1417–1424. Matthews, S., Power, C., & Stansfeld, S. (2001). Psychological distress and work and home roles: a focus on socioeconomic differences in distress. Psychological Medicine, 31, 725–736. Matthews, S., Stansfeld, S., & Power, C. (1999). Social support at age 33: the influence of gender, employment status and social class. Social Science and Medicine, 49(1), 133–142. Maughan, B., & Lindelow, M. (1997). Secular change in psychosocial risks: the case of teenage motherhood. Psychological Medicine, 27(5), 1129–1144. Mayfield, D., McLeod, G., & Hall, P. (1974). The CAGE questionnaire: validation of a new alcoholism screening instrument. American Journal of Psychiatry, 131, 112–1123. McLanahan, S., & Adams, J. (1989). The effects of children on adults’ psychological well being: 1957–1976. Social Forces, 68(1), 124–146. Miech, R. A., Caspi, A., Moffitt, T. E., Wright, B. R. E., & Silva, P. A. (1999). Low socioeconomic status and mental disorders: a longitudinal study of selection and causation during young adulthood. American Journal of Sociology, 104, 1096–1131. Muntaner, C., & Lynch, J. (1999). Income inequality, social cohesion, and class relations: a critique of Wilkinson’s neoDurkheimian research program. International Journal of Health Services, 29(1), 59–81. Offord, D. R., Boyle, M. H., Racine, Y. A., Fleming, J. E., Cadman, D. T., Blum, H. M., Byrne, C., Links, P. S., Lipman, E. L., & MacMillan, H. L. (1992). Outcome, prognosis, and risk in a longitudinal follow-up study. Journal of the American Academy of Child Adolescence Psychiatry, 31(5), 916–923. Oxman, T. E., Berkman, L. F., Kasl, S., Freeman, D. H., & Barrett, J. (1992). Social support and depressive symptoms in the elderly. American Journal of Epidemiology, 135(4), 356–368. Paykel, E. S., Myers, J. K., Dienelt, M. N., Klerman, G. L., Lindenthal, J. J., & Pepper, M. P. (1969). Life events and depression. A controlled study. Archives of General Psychiatry, 21(6), 753–760. Pearlin, L. I., & Johnson, J. S. (1977). Marital status, life strains and depressions. American Sociological Review, 42, 704–715. Power, C., & Manor, O. (1992). Explaining social class differences in psychological health among young adults: a longitudinal perspective. Socical Psychiatry and Psychiatric Epidemiology, 27, 284–291. Power, C., Manor, O., & Fox, A. J. (1991). Health and class: the early years. London: Chapman Hall. Power, C., & Matthews, S. (1997). Origins of health inequalities in a national population sample. Lancet, 350, 1584–1589. Quinton, D., & Rutter, M. (1988). Parenting Breakdown: the making and breaking of inter-generational links. Aldershot: Avebury.

2004

C. Power et al. / Social Science & Medicine 55 (2002) 1989–2004

Robins, L. N. (1978). Sturdy childhood predictors of adult antisocial behaviour: replications from longitudinal studies. Psychological Medicine, 8, 611–622. Robins, L. N., & Price, R. K. (1991). Adult disorders predicted by childhood conduct problems: results from the NIMH Epidemiologic Catchment Area project. Psychiatry, 54(2), 116–132. Rodgers, B., Pickles, A., Power, C., Collishaw, S., & Maughan, B. (1999). Validity of the Malaise Inventory in general population samples. Social Psychiatry and Psychiatric Epidemiology, 34(6), 333–341. Rodgers, B. (1990). Influences of early life and recent factors on affective disorder in women: an exploration of vulnerability models. In L. Robins, & M. Rutter (Eds.), Straight and devious pathways from childhood to adulthood (pp. 314–327). Cambridge: Cambridge University Press. Rodgers, B. (1991). Models of stress, vulnerability and affective disorder. Journal of Affective Disorders, 21(1), 1–13. Rodgers, B. (1991b). Socio-economic status, employment and neurosis. Social Psychiatry and Psychiatric Epidemiology, 26, 104–114. Rodgers, B., Korten, A. E., Jorm, A. F., Jacomb, P. A., Christensen, H., & Henderson, A. S. (2000a). Non-linear relationships in associations of depression and anxiety with alcohol use. Psychological Medicine, 30, 421–432. Rodgers, B., & Mann, S. L. (1993). Re-thinking the analysis of intergenerational social mobility: a comment on John W Fox’s ‘‘Social class, mental illness, and social mobility’’. Journal of Health and Social Behavior, 34, 165–172. Rutter, M. (1967). A children’s behaviour questionnaire for completion by teachers. Journal of Child Psychology and Psychiatry, 8, 1–11. Rutter, M., Quinton, D., & Hill, J. (1990). Adult outcome of institution-reared children: males and females compared. In L. Robins, & M. Rutter (Eds.), Straight and devious pathways from childhood to adulthood (pp. 135–157). Cambridge: Cambridge University Press. Rutter, M., Tizard, J., & Whitmore, K. (1970). Education, health and behaviour. London: Longman. Stansfeld, S. A., Fuhrer, R., & Shipley, M. (1998). Types of social support as predictors of psychiatric morbidity in a cohort of British civil servants Whitehall II Study. Psychological Medicine, 28, 881–892. Stansfeld, S. A., Fuhrer, R., Shipley, M. J., & Marmot, M. G. (1999b). Work characteristics predict psychiatric disorder: prospective results from the Whitehall II study. Occupational Environmental Medicine, 56(5), 302–307.

Stansfeld, S. A., Head, J., & Marmot, M. G. (1998). Explaining social class differences in depression and well being. Social Psychiatry and Psychiatric Epidemiology, 33, 1–9. Timms, D. (1996a). Social mobility and mental health in a Swedish cohort. Social Psychiatry and Psychiatric Epidemiology, 31(1), 38–48. Tremblay, R. E., Masse, B., Perron, D., LeBlanc, M., Schwartzman, A. E., & Ledingham, J. E. (1992). Early disruptive behaviour, poor school achievement, delinquent behaviour, and delinquent personality: longitudinal analyses. Journal of Consulting and Clinical Psychology, 60(1), 64–72. Turner, R. J., & Marino, F. (1994). Social support and social structure: a descriptive epidemiology. Journal of Health and Social Behavior, 35(3), 193–212. van de Mheen, H., Stronks, K., Schrijvers, C. T. M., & Mackenbach, J. P. (1999). The influence of adult ill health on occupational class mobility and mobility out of and into employment in The Netherlands. Social Science and Medicine, 49, 509–518. Verhulst, F. C., & van der Ende, J. (1992). Six-year developmental course of internalizing and externalizing problem behaviors. Journal of the American Academy of Child Adolescence Psychiatry, 31(5), 924–931. Wall, T. D., Bolden, R. I., Borrill, C. S., Carter, A. J., Golya, D. A., Shapiro, D. A, West, M. A., Hardy, G. E., Haynes, C. E., & Rick, J. E. (1997). Minor psychiatric disorder in NHS trust staff: occupational and gender differences. British Journal of Psychiatry, 171, 519–523. Wax, Y. (1992). Collinearity diagnostics for a relative risk regression analysis. Statistics in Medicine, 11, 1273–1287. Weich, S., & Lewis, G. (1998). Material standard of living, social class, and the prevalence of the common mental disorders in Great Britain. Journal of Epidemiological Community Health, 52(1), 8–14. Weich, S., Sloggett, A., & Lewis, G. (1998). Social roles and gender difference in the prevalence of common mental disorders. British Journal of Psychiatry, 173, 489–493. Weissman, M. M., Leaf, P. J., & Bruce, M. L. (1987). Single parent women. A community study. Social Psychiatry, 22(1), 29–36. West, P. (1991). Rethinking the health selection explanation for health inequalities. Social Science and Medicine, 32(4), 373–384. Wilkinson, R. G. (1997). Health inequalities: relative or absolute material standards? British Medical Journal, 314, 591–595.